docker-compose.yml

## Compose runtime parametrization through env vars
# if NST_HOST_MOUNT is set to a valid path, then the container will mount that path

version: '3'

services:

  # mkdir art
  # NST_HOST_MOUNT="$PWD/art" docker-compose up
  nst_demo:
    image: boromir674/neural-style-transfer:1.0.2
    volumes:
      - ${NST_HOST_MOUNT:-./nst-algo-output}:/app/demo-output
    command: [ "demo" ]


## NST Service, where the Pretrained Model Weights are made available to it (so that
# the code can read it), at build time (at docker build)
  # nst-service:
  #   build:
  #     context: ./
  #     target: prod_ready
  #     dockerfile: Dockerfile
  #     # at build time we can override the default location (in host) from where the builder tries to copy the to image    
  #     args:
  #       - IMAGE_MODEL=./pretrained_model_bundle/imagenet-vgg-verydeep-19.mat

    # ports:
    #   - "5000:5000"
    # volumes:
    #   - ./gen-images:/app/gen-images

    # networks:
    #   - nst-network



### OPT 2
## NST Service, where the Pretrained Model Weights are made available to it (so that
# the code can read it) at runtime (at docker run -e).

# So, at runtime, when we sping up the container, we mount a directory from local
# filesystem, where the model weights are stored, to a directory inside the container.

# We also make sure that at runtime we set the AA_VGG_19 variable to the path to the
# model weights inside the container's filesystem. That way, the code can read the
# model weights from the container's filesystem.

# Note, that changes in mounted Volumes have two-way effectL local and in-container
# In other words as, the directory of the model weights is in a "live" mode.
# That might be an unwanting side-effect, but it is not a problem for this project.

  # nst-service:
  #   build:
  #     context: .
  #     target: prod_install
  #     dockerfile: Dockerfile
  #   environment:
  #     AA_VGG_19: pretrained_model_bundle/imagenet-vgg-verydeep-19.mat
  #   volumes:
  #     - ./pretrained_model_bundle:/app/pretrained_model_bundle


  # mkdir art && NST_HOST_MOUNT="$PWD/art" docker-compose up
  # nst_demo_dev:
  #   build:
  #     context: ./
  #     # target: default_with_demo
  #     dockerfile: Dockerfile
  # # OR just build: .

  #   volumes:
  #     - ${NST_HOST_MOUNT:-./nst-algo-output}:/app/demo-output